Google is aware its AI sometimes won't answer who won the 2020 election: 'We're looking into this'
THE SO WHAT
Google’s AI Mode dodging a basic factual question about the 2020 US election highlights how safety filters can quietly degrade reliability on high-sensitivity topics. If you’re deploying AI in regulated or political contexts, you need explicit test suites for refusal patterns — not just accuracy — before you trust answers in production.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIExpert Judgement Can Be Easily Hacked by AI, Anthropic Study Finds
If 86% of participants follow Claude’s advice yet few recognize its influence, AI isn’t just a tool — it’s a hidden decision-maker in expert workflows. Any domain where “expert judgment” matters now needs explicit policies on AI consultation, disclosure, and second-level review.
Applied AISchwab Says AI Making Scams Harder to Spot
When Schwab’s risk head says AI is making scams harder to detect — including voice spoofing — that’s a signal that retail-facing security models are already behind. Any organization touching consumer money should be upgrading verification flows beyond voice and simple KYC this quarter.
Applied AIGoogle expands AI Search ads globally as core ad growth slows
Global rollout of AI Search ads as Alphabet’s core ad growth decelerates is the clearest sign that generative surfaces are becoming primary monetization real estate. Brands should expect more opaque auction dynamics and start testing how AI-shaped ad formats affect attribution and CAC now.
Applied AIMicrosoft recorded $10 billion in revenue from OpenAI
Pulling $10B in a quarter from commercial arrangements with OpenAI shows that frontier model partnerships can be P&L-scale, not just strategic. For enterprises, that concentration means your AI spend is indirectly feeding a small number of infra alliances — map that dependency explicitly.